S12 Bonus: The Microsoft File Trap: Moving Beyond Manual PowerPoint and Excel Workflows to Build an AI-Native "Consulting" Layer with Tim Lidman, Co-Founder & CEO of Clyde AI

Episode Summary

Executive Summary: Tim Lidman, co-founder and CEO of Clyde AI, explains how his consulting-software background and GenAI tools enabled him to build an AI-native collaboration platform that codifies consulting workflows. The episode covers Clyde’s rapid MVP, user-driven roadmap, evals and multi-LLM optimization, team-building, tech debt from AI-generated code, and the broader vision of democratizing high-end consulting through AI.

Main Topics: Origin of Clyde AI from consulting workflow pain points (Priority: 5/5): Lidman traces Clyde to his earlier startup ThinkTank, which helped consulting firms structure client workshops and engagements. While at Accenture, he saw consulting still run on manual documents and realized AI could codify the workflow. Rapid AI-native MVP development (Priority: 5/5): The team built an end-to-end private MVP in about three months using serverless Google Cloud, React, and heavy use of vibe coding. His co-founder built UX in Lovable and Claude, accelerating delivery and reducing design bottlenecks. Product maturation through user feedback and analytics (Priority: 5/5): After launch, Clyde shifted from founder intuition to user-driven prioritization. They use analytics plus AI to interrogate usage patterns conversationally and shape the roadmap based on what customers actually need. Engineering scale, evals, and model selection (Priority: 4/5): As usage diversified, the team implemented evals and test-driven workflows to manage quality. Clyde also uses a multi-LLM architecture to choose the right model per task and control token costs. AI-generated code debt and cleanup (Priority: 4/5): Lidman admits they overused AI code generation early on, creating tech debt and a gap between generated code and true understanding. They addressed it by rebuilding into a cleaner monorepo and migrating functionality over four weeks. Team culture and hiring philosophy (Priority: 4/5): Clyde operates with a very lean team and prioritizes AI-native, highly collaborative people who can work cross-functionally and interact with users. Trust and deep domain expertise were key in the founding team. Future of consulting and AI democratization (Priority: 5/5): Lidman sees Clyde as democratizing consulting, not eliminating it. He believes AI will extend expert-quality workflows to non-consultants while consulting firms specialize further, even as the AI market undergoes a correction.

Key Arguments: AI can codify consulting workflows that were previously manual, expensive, and heavily dependent on analysts and consultants. Deep domain expertise in both product and services creates an advantage because product teams and consulting teams usually think very differently. AI-native development dramatically reduces build time and cost, but it must be paired with disciplined engineering practices to avoid tech debt. A user-centered roadmap is more reliable than founder intuition once the product reaches real customers and varied use cases. Evals and structured testing are necessary because AI products behave variably across prompts and need systematic quality control at scale. Multi-LLM routing helps manage token scarcity and cost by selecting the cheapest capable model for each task. The broader impact of Clyde is democratizing access to high-end consulting-like outcomes for people outside elite firms. AI will not simply replace SaaS or consulting; instead, it will change how users interact with software and how service industries specialize.

Data Points: MVP build time: 3 months - Private MVP of Clyde was built end-to-end in about three months. Prior product build cost: $2 million - ThinkTank, the earlier collaboration platform, cost about $2 million to build over 18 months. MVP build cost: $250,000 - Clyde’s MVP was built for roughly $250K. Founding team size: 3 people - The company started with a founding group of three. Current team size: 6 people - Lidman says Clyde is operating with a very lean six-person team. Commercial launch timing: April of this year - Clyde commercially launched in April, about two months before the interview. Sign-ups: 1,700 users - He says about 1,700 people have signed up after launch. AI-generated code output: tens of thousands of lines - The team created tech debt by generating huge volumes of code too quickly. Cleanup migration window: 4 weeks - They spent about four weeks rebuilding into a new monorepo and migrating functionality. Prior startup acquisition year: 2021 - Accenture bought ThinkTank in 2021. Post-merger role duration: about 4 years - He stayed on at Accenture as a partner after the acquisition.

Pivotal Quotes: "We got ahead of ourselves on using AI to generate just insane amounts of code." — Tim Lidman: He explains how rapid AI-assisted development created technical debt. "We’re basically writing eval scripts, running them, and then gap analyzing, adjusting." — Tim Lidman: He describes how Clyde now uses evals and test-driven methods to improve AI product quality. "I think we’re democratizing consulting in a way where the existing consulting industry are going to find ways to deeply specialize." — Tim Lidman: He frames Clyde’s long-term impact on the consulting industry and broader access to expertise.

Implications: The episode suggests AI-native products can compress startup timelines, but sustainable scale requires evals, model optimization, and disciplined code management. More broadly, AI may broaden access to expert workflows, not just automate existing SaaS or consulting.